PUBLICATIONS

Overview

Meier, L. J. (2025): Medical and Clinical Ethics of AI. In: A Companion to Applied Philosophy of AI, edited by M. Hähnel and R. Müller. West Sussex: Wiley Blackwell [forthcoming]

Popular Literature

Abstracts

Meier, L. J. (2025): Medical and Clinical Ethics of AI. In: A Companion to Applied Philosophy of AI, edited by M. Hähnel and R. Müller. West Sussex: Wiley Blackwell [forthcoming]

Machine intelligence is being employed for an increasing number of tasks in the healthcare sector. The more encompassing these tasks are becoming, the likelier it will be that, besides working with empirical data, algorithms are also coming into contact with normative aspects. This is a consequence of the fact that doing medicine is intimately intertwined with the taking of value judgments. Will the automation of healthcare therefore result in a new – this time algorithmic – form of medical paternalism? I argue that two design requirements could mitigate this tendency: embedding moral theories into medical AI and equipping it with input mechanisms for individual and situational preferences and values. I analyse how well consequentialist, deontological, and virtue-ethical theories are suited for algorithmic implementation and propose principlism as a compromise solution. Finally, I sketch how value input can be realised on this basis.

Under which conditions may physicians who are requested to treat permanently unconscious patients refuse to do so? Wasserman et al. (2023) maintain that refusals on the basis of supposed futility are unethical as they amount to passing off personal value judgments as medical expertise. Instead, unwillingness to carry out an intervention should be framed as conscientious objection. I argue that referring to futility with regard to a patient’s presumed quality of life is appropriate if – and only if – a total and irreversible loss of consciousness has been diagnosed with adequate certainty. The diagnosis turns a graded notion into a binary affair: conscious awareness is a necessary condition of any degree of quality of life to be meaningfully ascribed to an organism. Consequently, the complete absence of consciousness precludes such an ascription; further treatment is then indeed futile from a quality-of-life standpoint.

During the Cold-War era, intelligence agencies in both the Eastern and Western blocs set up secret experiments to devise methods of mind control and brainwashing in order to gain an advantage over the enemy. The most notorious of the these endeavours was the CIA’s illegal MKUltra programme, which ran from from 1953 to 1973. In its quest for a truth serum that could be employed to force confessions from subjects who are not cooperating and the desire to be able to manipulate people’s behaviour, the CIA experimented on thousands of subjects – partly without their knowledge or consent and often with disastrous consequences. In this paper, I scrutinise the MKUltra programme, investigating some of the experiments that it comprised. I also map the programme’s aims onto the latest technology available for interfering with the brain – so-called brain-computer interfaces – to determine whether one could, in principle, use these novel devices for the purpose of mind control. Finally, I am looking for indications that stakeholders may actually plan to employ the technology to achieve what had been beyond technological reach during the Cold War. I conclude that brain-computer interfaces could indeed be used to realise some of the original goals; and that an interest in mind control still prevails.

Integrating artificial intelligence (AI) into critical domains such as healthcare holds immense promise. Nevertheless, significant challenges must be addressed to avoid harm, promote the well-being of individuals and societies, and ensure ethically sound and socially just technology development. Innovative approaches like Embedded Ethics, which refers to integrating ethics and social science into technology development based on interdisciplinary collaboration, are emerging to address issues of bias, transparency, misrepresentation, and more. This paper aims to develop this approach further to enable future projects to effectively deploy it. Based on the practical experience of using ethics and social science methodology in interdisciplinary AI-related healthcare consortia, this paper presents several methods that have proven helpful for embedding ethical and social science analysis and inquiry. They include (1) stakeholder analyses, (2) literature reviews, (3) ethnographic approaches, (4) peer-to-peer interviews, (5) focus groups, (6) interviews with affected groups and external stakeholders, (7) bias analyses, (8) workshops, and (9) interdisciplinary results dissemination. We believe that applying Embedded Ethics offers a pathway to stimulate reflexivity, proactively anticipate social and ethical concerns, and foster interdisciplinary inquiry into such concerns at every stage of technology development. This approach can help shape responsible, inclusive, and ethically aware technology innovation in healthcare and beyond.

Automatic Patient Preference Predictors are algorithms that use statistically predictive demographic key characteristics to infer an individual’s treatment preferences in health care. Lately, this initial proposal was expanded into a personalised version: instead of relying on demographic data, large language models are to extract values from a variety of sources generated by individual patients. I sketch two potential problems with the two arguably most transformative types of data source on which such an algorithm would be based: the difficulty of differentiating between general and idiosyncratic statements in online activities, and the tension between incentivising users and the faithfulness of the preferences so elicited. The accuracy of personalised predictor algorithms might be much lower than currently expected.

During the COVID-19 pandemic, national triage guidelines were developed to address the anticipated shortage of life-saving resources. Rationing and triage imply that, in addition to individual patient interests, population health must also be considered. The paper analyses how triage protocols can translate abstract theories of distributive justice into concrete material and procedural criteria for rationing intensive-care resources. We reconstruct the development and implementation of a rationing protocol at a German university hospital, describing the ethical challenge of triage, clarifying the aspirational norms, and summarising specific norms of fair triage and allocation for developing and implementing an institutional policy and practice model. We reflect on how critical topics are seen by clinicians and which interventions helped to manage the perceived burdens of triage dilemmas. The article seeks to inform the debate about triage concepts and policies to ensure the best possible treatment and fair allocation of resources as well as to help protect both patients and professionals in worst-case scenarios.

In their Target Article, Rahimzadeh et al. (2023) discuss the virtues and vices of employing ChatGPT in ethics education for healthcare professionals. To this end, they confront the chatbot with a moral dilemma and analyse its response. In interpreting the case, ChatGPT relies on Beauchamp and Childress’ four prima-facie principles: beneficence, non-maleficence, respect for patient autonomy, and justice. While the chatbot’s output appears admirable at first sight, it is worth taking a closer look: ChatGPT not only misses the point when applying the principle of non-maleficence; its response also fails, in several places, to honour patient autonomy – a flaw that should be taken seriously if large language models are to be employed in ethics education. I therefore subject ChatGPT’s reply to detailed scrutiny and point out where it went astray.

Lockean views of personal identity maintain that we are essentially persons who persist diachronically by virtue of being psychologically continuous with our former selves. In this article, I present a novel objection to this variant of psychological accounts, which is based on neurophysiological characteristics of the brain. While the mental states that constitute said psychological continuity reside in the cerebral hemispheres, so that for the former to persist only the upper brain must remain intact, being conscious additionally requires that a structure originating in the brainstem – the ascending reticular activating system – be functional. Hence, there can be situations in which even small brainstem lesions render individuals irreversibly comatose and thus forever preclude access to their mental states, while the neural correlates of the states themselves are retained. In these situations, Lockeans are forced to regard as fulfilled their criterion of diachronic persistence since psychological continuity, as they construe it, is not disrupted. Deeming an entity that is never again going to have any mental experiences to be a person, however, is an untenable position for a psychological account to adopt. In their current form, Lockean views of personal identity are therefore incompatible with human neurophysiology.

Zuk et al. (2023) examined researchers’ views on how deep brain stimulation may impact patients’ personality, mood, and behaviour (PMB). The team found that experts vary substantially in the notion of personality they employ. However, despite noting the lack of conceptual precision, no attempt was made at scientifically defining any of the involved concepts, so that the results of the different interviews remain largely incommensurable.
In this comment, I am doing the interpretative work that the authors should have undertaken following the descriptive part of the paper: first, disentangling the PMB cluster by defining what exactly constitutes personality, mood, and behaviour; and secondly, conceptualising the notion of personality change. I am arguing that what sets personality changes apart from other modifications is diachronic persistence.

Machine intelligence already helps medical staff with a number of tasks. Ethical decision-making, however, has not been handed over to computers. In this proof-of-concept study, we show how an algorithm based on Beauchamp and Childress’ prima-facie principles could be employed to advise on a range of moral dilemma situations that occur in medical institutions. We explain why we chose fuzzy cognitive maps to set up the advisory system and how we utilized machine learning to train it. We report on the difficult task of operationalizing the principles of beneficence, non-maleficence and patient autonomy, and describe how we selected suitable input parameters that we extracted from a training dataset of clinical cases. The first performance results are promising, but an algorithmic approach to ethics also comes with several weaknesses and limitations. Should one really entrust the sensitive domain of clinical ethics to machine intelligence?

Although machine intelligence is increasingly employed in healthcare, the realm of decision-making in medical ethics remains largely unexplored from a technical perspective. We propose an approach based on fuzzy cognitive maps (FCMs), which builds on Beauchamp and Childress’ prima-facie principles. The FCM’s weights are optimized using a genetic algorithm to provide recommendations regarding the initiation, continuation, or withdrawal of medical treatment. The resulting model approximates the answers provided by our team of medical ethicists fairly well and offers a high degree of interpretability. Possible applications of such a system include informal guidance on medical ethics dilemmas as well as educational purposes.

Can machine intelligence do clinical ethics? And if so, would applying it to actual medical cases be desirable? In a recent target article (Meier et al. 2022), we described the piloting of our advisory algorithm METHAD. Here, we reply to commentaries published in response to our project. The commentaries fall into two broad categories: concrete criticism that concerns the development of METHAD; and the more general question as to whether one should employ decision-support systems of this kind—the debate we set out to ignite with our target article.

The COVID-19 pandemic has been overwhelming public health-care systems around the world. With demand exceeding the availability of medical resources in several regions, hospitals have been forced to invoke triage. To ensure that this difficult task proceeds in a fair and organised manner, governments scrambled experts to draft triage guidelines under enormous time pressure. Although there are similarities between the documents, they vary considerably in how much weight their respective authors place on the different criteria that they propose. Since most of the recommendations do not come with ethical justifications, analysing them requires that one traces back these criteria to their underlying theories of distributive justice. In the literature, COVID-19 triage has been portrayed as a value conflict solely between utilitarian and egalitarian elements. While these two accounts are indeed the main antipodes, I shall show that in fact all four classic theories of distributive justice are involved: utilitarianism, egalitarianism, libertarianism, and communitarianism. Detecting these in the documents and classifying the suggested criteria accordingly enables one to understand the balancing between the different approaches to distributive justice—which is crucial for both managing the current pandemic and in preparation for the next global health crisis.

Fifty years have passed since brain death was first proposed as a criterion of death. Its advocates believe that with the destruction of the brain, integrated functioning ceases irreversibly, somatic unity dissolves, and the organism turns into a corpse. In this article, I put forward two objections against this assertion. First, I draw parallels between brain death and other pathological conditions and argue that whenever one regards the absence or the artificial replacement of a certain function in these pathological conditions as compatible with organismic unity, then one equally ought to tolerate that function’s loss or replacement in brain death. Second, I show that the neurological criterion faces an additional problem that is only coming to light as life-supporting technology improves: the growing sophistication of the latter gives rise to a dangerous decoupling of the actual performance of a vital function from the retention of neurological control over it. Half a century after its introduction, the neurological criterion is facing the same fate as its cardiopulmonary predecessor.

Good physical experiments conform to the basic methodological standards of experimental design: they are objective, reliable, and valid. But is this also true of thought experiments? Especially problems of personal identity have engendered hypothetical scenarios that are very distant from the actual world. These imagined situations have been conspicuously ineffective at resolving conflicting intuitions and deciding between the different accounts of personal identity. Using prominent examples from the literature, I argue that this is due to many of these thought experiments not adhering to the methodological standards that guide experimental design in nearly all other disciplines. I also show how empirically unwarranted background assumptions about human physiology render some of the hypothetical scenarios that are employed in the debate about personal identity highly misleading.

Objectives: Simulation and evaluation of a prioritization protocol at a German university hospital using a convergent parallel mixed methods design.

Design: Prospective single-center cohort study with a quantitative analysis of ICU patients and qualitative content analysis of two focus groups with intensivists.

Setting: Five ICUs of internal medicine and anesthesiology at a German university hospital.

Patients: Adult critically ill ICU patients ( n = 53).

Interventions: After training the attending senior ICU physicians ( n = 13) in rationing, an impending ICU congestion was simulated. All ICU patients were rated according to their likelihood to survive their acute illness (good-moderate-unfavorable). From each ICU, the two patients with the most unfavorable prognosis ( n = 10) were evaluated by five prioritization teams for triage.

Measurements and main results: Patients nominated for prioritization visit ( n = 10) had higher Sequential Organ Failure Assessment scores and already a longer stay at the hospital and on the ICU compared with the other patients. The order within this worst prognosis group was not congruent between the five teams. However, an in-hospital mortality of 80% confirmed the reasonable match with the lowest predicted probability of survival. Qualitative data highlighted the tremendous burden of triage and the need for a team-based consensus-oriented decision-making approach to ensure best possible care and to support professionals. Transparent communication within the teams, the hospital, and to the public was seen as essential for prioritization implementation.

Conclusions: To mitigate potential bias and to reduce the emotional burden of triage, a consensus-oriented, interdisciplinary, and collaborative approach should be implemented. Prognostic comparative assessment by intensivists is feasible. The combination of long-term ICU stay and consistently high Sequential Organ Failure Assessment scores resulted in a greater risk for triage in patients. It remains challenging to reliably differentiate between patients with very low chances to survive and requires further conceptual and empirical research.

When an individual is comatose while parts of her brain remain functional, the question arises as to whether any mental characteristics are still associated with this brain, that is, whether the person still exists. Settling this uncertainty requires that one becomes clear about two issues: the type of functional loss that is associated with the respective profile of brain damage and the persistence conditions of persons. Medical case studies can answer the former question, but they are not concerned with the latter. Conversely, in the philosophical literature, various accounts of personal identity are discussed, but usually detached from any empirical basis. Only uniting the two debates and interpreting the real-life configurations of brain damage through the lens of the philosophical concepts enables one to make an informed judgment regarding the persistence of comatose persons. Especially challenging are cases in which three mental characteristics that normally occur together – wakefulness, awareness and memory storage – come apart. These shall be the focus of this paper.

When does a human being cease to exist? For millennia, the answer to this question had remained largely unchanged: death had been diagnosed when heartbeat and breathing were permanently absent. Only comparatively recently, in the 1950s, rapid developments in intensive-care medicine called into question this widely accepted criterion. What had previously been deemed a permanent cessation of vital functions suddenly became reversible. A new criterion of death was needed. It was suggested that the destruction of the brain could indicate the death of the organism in the presence of external life support. Soon the so-called brain death became the new worldwide standard. In recent years, however, doubts about this neurological criterion have been growing. Is brain death really our death? This is the question that this thesis seeks to answer. To this end, we shall connect the medical debate about the definition of death to the philosophical debate about personal identity. While we will find that the destruction of its brain does in fact not correspond to an organism’s death, we shall also ask whether the assumption that we are essentially organisms is correct. May brain death be the ceasing to exist of a different entity? Substituting clinical case reports and considerations about human physiology for the use of thought experiments, the thesis takes a novel and philosophically unconventional approach to the problem of what we essentially are. We shall analyse various pathological conditions and their respective effects on the bodily and mental characteristics of our existence. We will conclude that brain death is indeed our death – but for reasons entirely different from those cited in the original justification of this criterion.

By appealing to the similarity between pre-vital and post-mortem nonexistence, Lucretius famously tried to show that our anxiety about death was irrational. His so-called Symmetry Argument has been attacked in various ways, but all of these strategies are themselves problematic. In this paper, I propose a new approach to undermining the argument: when Parfit’s distinction between identity and what matters is applied, not diachronically (as he uses it) but across possible worlds, the alleged symmetry can be broken. Although the pre-vital and posthumous time spans that we could have experienced are indeed analogous with respect to our identity, they are not analogous with respect to psychological continuity, which forms the basis of prudential concern. Lucretius even anticipated the Parfitian distinction. He did not, however, notice the significance that it has for his Symmetry Argument.

Jakob Michael Reinhold Lenz’ 1776 tragicomedy The Soldiers is one of the most important works of the German Sturm und Drang movement. Turning away from the Aristotelian concept of drama, Lenz uses the writing of letters to indirectly characterise the play’s protagonists. I analyse his revolutionary methodology that places emphasis on the process of composition rather than – as was customary in his times – focusing on the actual content of what has been written.